Chemical Accuracy in Computational Prediction of PFAS–Cyclodextrin Binding: A Multilevel SQM/QM Study With Experimental Validation

Abstract

Achieving chemical accuracy in the prediction of host–guest binding free energies remains a central challenge in computational chemistry, particularly for flexible supramolecular systems in aqueous environments. Here, we demonstrate that a multilevel CREST/CENSO SQM/QM workflow based on Grimme's r²SCAN‐3c methodology can predict β‐cyclodextrin binding free energies for perfluoroalkyl carboxylic acids (PF4–PF8) with near‐quantitative agreement with experiment. Calculated binding free energies were validated against isothermal titration calorimetry (ITC) measurements, yielding a mean absolute error of 0.86 kcal mol⁻¹ and thereby achieving chemical accuracy. Both computational and experimental results reveal a pronounced chain‐length dependence, with binding affinity decreasing toward shorter PFAS. Energy decomposition analysis shows that selectivity is governed primarily by electronic interactions, dominated by dispersion along the fluorocarbon chain. Noncovalent interaction analysis further identifies carboxylate hydrogen bonding at the primary rim of β‐cyclodextrin as a key anchoring interaction. These results establish an efficient and robust quantum‐mechanical framework for predicting PFAS–cyclodextrin complexation and provide a rational basis for the computational screening and design of cyclodextrin‐based sorbents for PFAS remediation.

Accurate prediction of β‐cyclodextrin complexation with PFAS (per‐ and polyfluoroalkyl substances, PF4–PF8) is achieved using a multilevel CREST/CENSO SQM/QM workflow based on r²SCAN‐3c. Binding free energies agree with isothermal titration calorimetry (ITC) with a mean absolute error of 0.86 kcal mol⁻¹, providing a reliable quantum‐mechanical framework for screening cyclodextrin‐based PFAS sorbents.

1. Introduction

Poly‐ and perfluoroalkyl substances (PFAS) have garnered global attention due to their widespread industrial and consumer uses with extraordinary chemical stability [1, 2]. These synthetic fluorinated compounds are characterized by an amphiphilic structure comprising a fully or partially fluorinated hydrophobic alkyl chain and a polar, often anionic, head group—typically a carboxylate or sulfonate [3]. Their exceptional thermal and chemical inertness arises from the carbon–fluorine (C—F) bond, one of the strongest covalent bonds known in organic chemistry [2]. This has enabled PFAS to persist and accumulate extensively across environmental compartments, earning PFAS the widely recognized designation of “forever chemicals” [4, 5]. Further, environmental and health implications of PFAS are becoming ever‐increasingly demonstrated, whereby PFAS generally do not metabolize but rather bioaccumulate [2, 5, 6]. Notably, PFAS exposure is associated with immunotoxicity, endocrine dysfunction, and hepatotoxicity, while epidemiological studies have linked human exposure to long‐chain PFAS like PFOA and PFOS to a high incidence of kidney cancer, testicular cancer, high cholesterol, thyroid disease, ulcerative colitis, neurotoxicity, and pregnancy‐induced hypertension [4, 5, 7].

Remediation of PFAS‐contaminated water and environmental matrices continues to pose a substantial technical challenge, with conventional water and wastewater treatments being largely ineffective [3, 7]. Furthermore, degradation also shows distinct problems as a remediation strategy, as even advanced oxidation or other chemical treatment methods often achieve only partial decomposition [5], while concern is that the breakdown of longer‐chain PFAS precursors generates stable shorter‐chain intermediates (e.g., PFPeA, PFHxA, and PFHpA), which often are more mobile, more difficult to remove, and in some cases, more toxic than their parent compounds [8, 9]. Adsorption is hence the most widely studied and implemented approach for PFAS removal, but its efficacy is strongly dependent on PFAS chain length [5, 7], whereby long‐chain PFAS are primarily removed through hydrophobic interactions [5, 10], while short‐chain PFAS exhibit hydrophilicity and mobility, which leads to limited adsorption thereby complicating their removal [5]. Cyclodextrins (CDs) have attracted significant attention in potential sorbent systems for PFAS remediation. These cyclic oligosaccharides possess a ring‐shaped structure with a hydrophilic exterior and a hydrophobic cavity [11, 12], forming strong host–guest inclusion complexes with PFAS [13, 14]. This capability has prompted the design of CD‐based adsorbents that effectively remove PFAS from contaminated water [11]. Despite promising results, many aspects of CD‐PFAS complexation remain unclear [11], and understanding the formation and dynamics of these inclusion complexes is crucial for developing sorbents that can effectively remove PFAS, particularly short‐chain derivatives [15].

While CD–PFAS complexes have been frequently observed using NMR measurements [11, 16, 17], there has only been one published isothermal titration calorimetry (ITC) measurement observing the interaction between β‐CD and PFOA [18]. Although NMR titrations give good indication regarding chemical groups associated with interactions, the host–guest energies are associated with the thermodynamics determined through ITC measurements. This is an underutilized method for understanding PFAS binding. ITC is a straightforward and elegant method employed for the direct determination of molecular interactions whereby the heat associated from interactions is the thermodynamic observable which relates to enthalpy [19, 20]. Further, ITC not only provides the enthalpic (ΔH°) contributions, but also the entropic contributions (ΔS°) and Gibbs free energy (ΔG°) associated with interactions along with binding association constants (K _d or K _a) and stoichiometry (n). Combining ITC with chemical computational methods could illuminate fundamental thermodynamical processes associated with complexation, hence allowing for rational design of new systems.

The statistical assessment of modeling of proteins and ligands (SAMPL) blind prediction challenges provides rigorous benchmarking frameworks that validate and improve state‐of‐the‐art computational methods [21]. For cucurbituril (CB) systems (SAMPL8), even sophisticated multilevel quantum mechanical (QM) protocols achieved a moderate accuracy, measured by a mean absolute deviation (MAD), of 1.82 kcal mol⁻¹ [21]. In contrast, the multilevel QM refinement of Jameel and Stein reached the targeted thermochemical accuracy (<1 kcal mol⁻¹) for β‐CD systems [21]. Meanwhile, force field‐based methods, while computationally efficient, are problematic because the complexity of molecular systems requires vast parametrizations [22] with specialized force fields often necessary for nonstandard binding situations [21]. The development of general, applicable force fields remains difficult due to the necessity of fitting numerous atom‐pair‐specific parameters to experimental or first‐principle data [22]. Historically, finding an adequate balance between computational cost and accuracy remains a bottleneck in CD chemistry [23]. Although the number of works published using quantum mechanics on CD host–guest systems has increased, it is still relatively low compared to contributions using molecular dynamics and molecular mechanics levels [23].

This gap is particularly striking given recent methodological advances that have made chemical accuracy achievable for flexible supramolecular systems without prohibitive computational cost [21, 24]. Notably, Grimme's multilevel CREST/CENSO workflow combines efficient conformational sampling, typically performed using semiempirical extended tight‐binding methods (GFN‐xTB) or the generalized force field (GFN‐FF) [25], with subsequent refinement using robust composite density functional theory (DFT) methods like r²SCAN‐3c [26, 27]. This validated QM approach achieves performance comparable to methods requiring extensive sampling, requiring 2–3 orders of magnitude fewer structures [21]. The successful application and validation of these state‐of‐the‐art QM protocols [21] enable their use for the systematic study of challenging complexation reactions, such as PFAS complexation with β‐CD across a homologous chain‐length series. Accurate prediction of host–guest binding free energies while achieving “chemical accuracy” (errors below ~1 kcal mol⁻¹) remains one of the grand challenges in computational chemistry [21]. As such, the primary objective of the work presented here is to achieve effective CD–PFAS (host–guest) QM calculations for the determination of associated binding energies, which would be validated through experimental ITC analysis. This work presents the first systematic DFT‐level investigation of CD‐PFAS binding, applying Grimme's CREST/CENSO multilevel strategy to the homologous series perfluorobutanoic acid (PF4) through to perfluorooctanoic acid (PF8) with native β‐CD. Energy decomposition could reveal structure–affinity relationships based on the length of the alkyl‐tail, providing a foundation for computational screening of CDs and their derivatives for binding PFAS in rational remediation design.

2. Results and Discussion

2.1. Chain‐Length‐Dependent Binding Modes

Conformational optimization yielded distinct binding geometries across the PFAS series, with all complexes converging to stable 1:1 inclusion modes. The structural characterization reveals systematic trends in PFAS orientation and insertion depth (Figure 1).

FIGURE 1.

Lowest‐energy conformers of β‐CD complexes with perfluoroalkyl acids of varying chain length. (A) Side view and (B) bottom view from the primary rim, showing the H‐bonding by red dashed lines. PFAS: space‐filling (C: black, F: cyan, O: red); cyclodextrin: line representation.

Structural analysis reveals chain‐length‐dependent binding modes. The short‐chain PFAS (PF4, PF5) achieve complete insertion with carboxylate groups positioned near the narrower primary rim, maximizing host–guest contact. Longer‐chain PFAS (PF7, PF8) adopt partially protruding geometries with terminal —CF₃ groups extending beyond the secondary rim. This incomplete insertion for longer‐chain PFAS suggests a balance between maximizing cavity interactions and maintaining carboxylate solvation. The reverse orientation was also investigated for PF8 (Figure S1), with a ΔG _calc of −4.3 kcal mol⁻¹ calculated, indicating that the PFAS with COO⁻ oriented toward the primary rim of CD (ΔG _calc of −6.4 kcal mol⁻¹) is the dominant conformation.

The observed carboxylate‐to‐primary‐rim orientation contrasts with phenothiazine‐β‐CD complexes, where structural analyses indicate the phenothiazine moiety is located at the secondary face (the wider rim) of the CD host [21], reflecting distinct hydrophobic–hydrophilic profiles of these guest classes. This anionic‐guest binding mode, which targets the narrow primary rim (tail) containing the primary hydroxyl groups [28], aligns with computational studies acknowledging that the formation of such systems relies heavily on strong polar interactions, such as ionic hydrogen bonds [21]. The need for high‐level theory to accurately describe these interactions underscores the primary rim's crucial role in binding charged species, as simpler semiempirical methods often exhibit poor performance in describing the necessary polarization effects of these strong ionic bonds [21, 29, 30].

Interestingly, despite deeper insertion of shorter chains suggesting stronger geometric complementarity, the energetic analysis reveals an inverse relationship between insertion depth and binding affinity.

2.2. Energy Decomposition and Physical Interpretation

To understand the counterintuitive relationship between insertion depth and binding affinity, binding free energies were decomposed into their constituent contributions according to Equation (1):

$$\mathrm{\Delta }{G}_{\mathrm{calc}}\hspace{0.17em}=\hspace{0.17em}\mathrm{\Delta }E\hspace{0.17em}+\hspace{0.17em}\mathrm{\Delta }{G}_{\mathrm{solv}}\hspace{0.17em}+\hspace{0.17em}\mathrm{\Delta }{G}_{\mathrm{mRRHO}}$$ (1)

This partitioning isolates electronic interaction energies (ΔE), differential solvation penalties (ΔG _solv), and thermostatistical terms including entropy (ΔG _mRRHO), enabling identification of the dominant physical factors governing chain‐length selectivity (Figure 2).

FIGURE 2.

Energy decomposition analysis of β‐CD‐PFAS binding free energies. Stacked bars show individual contributions: electronic interaction energy ΔE (gray), solvation free energy ΔG _solv (orange), and thermostatistical contribution ΔG _mRRHO (blue). Total calculated binding free energies ΔG _calc (black line with values) demonstrate weakening toward shorter chains. All values reported in kcal mol⁻¹ at 298.15 K. Raw values are indicated in Table S1.

Energy decomposition reveals that electronic interactions (ΔE) are strongly favorable for all complexes and exhibit a systematic chain‐length dependence, varying by ~3 kcal mol⁻¹ across the series. The solvation term (ΔG _solv) systematically increases (becomes more positive/unfavorable) with decreasing chain length, ranging from ~26.5 kcal mol⁻¹ for PF8 to ~28.3 kcal mol⁻¹ for PF4. This ~1.8 kcal mol⁻¹ solvation penalty escalation reinforces the ~1.6 kcal mol⁻¹ reduction in total binding affinity (ΔG) from PF8 to PF4 [27, 31, 32].

The thermostatistical contribution (ΔG _mRRHO) decreases from +21.7 to +18.8 kcal mol⁻¹ across PF8‐PF4, partly offsetting the solvation trend and possibly reflecting a smaller entropy loss upon complexation for shorter guests; accurate calculation of this term is vital for determining the overall free energy of association [27, 31, 32].

This compensation pattern aligns with established principles in supramolecular chemistry, where desolvation penalties frequently oppose favorable gas‐phase electronic interactions [27, 31]. The critical role of implicit solvation models (ΔG _solv) in achieving chemical accuracy for flexible CD systems has been consistently demonstrated in multilevel QM workflows [21, 31], emphasizing that neglecting these continuum solvent contributions would lead to qualitatively incorrect predictions for cyclodextrin systems [28].

2.3. Experimental (ITC) Binding Energies

Initial ITC measurements allowed for appropriate experimental conditions (Table S2) and subsequent modeling based on the ITC ‘c’ value (Equation (2)), given optimal modeling where 50 < ‘c’ < 500 [33] dependent on the binding association (K), host molecule concentration (i.e., CD), and stoichiometric ratio (n). As such, experimental conditions were ascertained while considering the critical micelle concentration (CMC) of PFAS and solubility of CD in water. However, PF4 complexation with CD was not appropriate for ITC modeling, reaching lower limits of the software fitting due to weak binding even at the upper limits of CD solubility, with this low binding visible in the raw data of the corrected heat rates for all the titrations (Figure S2).

$$‘c’\hspace{0.17em}=\hspace{0.17em}n\hspace{0.17em}·\hspace{0.17em}{K}_{\mathrm{a}}\hspace{0.17em}·\hspace{0.17em}\left[\mathrm{CD}\right]\hspace{0.17em}\mathrm{or}\hspace{0.17em}n\hspace{0.17em}·\hspace{0.17em}\left[\mathrm{CD}\right]/K_{\mathrm{d}}$$ (2)

Binding energies determined through ITC results indicated a clear trend, being that the shorter‐chain PFAS result in lower binding affinities (K _a) and more positive enthalpic energies (ΔH°) associated with complexation to CD (Figure 3). Notably, there is an additional energy associated with relatively low PF8 and PF7 ratios (<0.3) to CD, corresponding with the previously reported 2:1 complexation of CD:PF8 [18] and similarly, 2:1 complexation of CD:PF7 [16]. As these 2:1 complexations have been described previously, we report only on the 1:1 binding as the focus of our work is to describe the host–guest binding energy and its dependence on the PFAS chain length. Further, despite our best efforts to increase the ‘c’ value for PF4‐CD complexation through concentration adjustments, binding of PF4 with CD was ultimately substantially lower than the limits of good acquisition and modeling of ITC results [33]. Nonetheless, it is noticeable that the enthalpy (ΔH°) of complexation with CD becomes less favorable with shorter‐chain PFAS (Table 1), resulting in a lower binding affinity constant (K _a), which then results in a more positive ΔG°. This indicates that a decrease in PFAS chain‐length results in a decrease in the thermodynamic driving force for complexation with CD. Nonetheless, ΔG° values indicate all PFAS observed here complex with CD spontaneously in an aqueous environment. With the principal factors involved in complexation being van der Waals and hydrophobic interactions, the obtained values of ΔH° and ΔS° lead to differentiation between these types of noncovalent forces involved, complexations where |ΔH| > |TΔS| indicate van der Waals interactions are the enthalpy‐driven dynamics with minor favorable or unfavorable entropies of interaction [34] while complexations with |ΔH| < |TΔS| indicating hydrophobic interactions and being entropy‐driven dynamics, where the entropy of the interaction is large and positive whereas the enthalpy of the process is small [35]. From this, it becomes apparent that CD complexation with PFAS presented here, regardless of the chain length, is enthalpy‐driven whereby van der Waals interactions are predominating with favorable entropy.

FIGURE 3.

ITC experimental results (circles) and modeled fit (solid line) for titrations of PFAS to β‐CD in distilled water.

TABLE 1.

ITC determined thermodynamic values for PFAS‐β‐CD complexation in distilled water. Values shown with 95% CIs.

	K a, M−1	n	ΔH°, kJ mol−1	ΔG°, kcal mol−1	ΔS°, J (mol K)−1	TΔS°, kJ mol−1
PF8	2.44 × 105 (±2.45 × 104)	1.00 (±0.01)	−23.03 (±0.22)	−7.35 (±0.06)	25.89 (±1.12)	7.72 (±0.33)
PF7	8.41 × 104 (±8.93 × 103)	0.95 (±0.01)	−21.71 (±0.31)	−6.72 (±0.06)	21.46 (±1.37)	6.40 (±0.41)
PF6	2.27 × 104 (±2.21 × 103)	0.96 (±0.01)	−18.12 (±0.26)	−5.94 (±0.06)	22.63 (±1.20)	6.74 (±0.36)
PF5	2.68 × 103 (±2.45 × 102)	1.01 (±0.01)	−12.85 (±0.20)	−4.71 (±0.06)	23.01 (±1.01)	6.86 (±0.30)
PF4	<1.00 × 103	NA	NA	NA	NA	NA

Results from ITC agree with global trends of the thermodynamic relationships for CD complexation with a guest molecule (for a 1:1 stoichiometry), whereby the structural and electronic features of the guest molecule can be rationalized, but not limited to, in terms of hydrophobicity and substituent(s) on aliphatic guests available for hydrogen bonding [36]. Further, the parts of the guest molecule where a change is induced due to its environment from CD complexation, such as solvation or hydrophobicity, contribute appreciably to the complexation thermodynamics, while CD also undergoes substantial changes such as desolvation [36]. While thermodynamics associated with CD–surfactant complexation is well known [37, 38], PFAS have yet to be scrutinized in the same manner, although notably PFAS employed in this study is akin to surfactants, pertaining to a hydrophobic tail and polar head group. Thereby, the understanding of thermodynamical processes associated with CD–surfactant interactions can illuminate the CD–PFAS complexation thermodynamics. Binding between CDs and surfactants is a complex and dynamic process, including dehydration of both the surfactant and CD, complexation of the surfactant into the CD cavity, hydration of the complex, establishment of equilibria, and potential supramolecular growth [37]. Notably for CD complexation, high‐energy water molecules are initially released, with these water molecules highly structured (and hence high energy) because they hydrate the hydrophobic tail of the surfactant (or PFAS in this study) as well as the CDs' hydrophobic and spatially confined cavity [39, 40], with this water release representing an entropy‐favored process and being a dominant driving force for complexation in aqueous environments [37]. However, here we show that although the entropy is favorable for PFAS complexation and likely promotes interaction, the complexation is indeed predominantly enthalpy driven (i.e., stabilized via van der Waals interactions). While small molecules represent simpler hydrophobic partitioning effects, surfactants and PFAS equivalents require cooperative alignment of their hydrophobic chains within the cavity of CD, often through a favorable orientation and position, while the structure of internalized water within the CD cavity is disrupted [38]. The amount of high‐energy water hydrating the hydrophobic part of the surfactant depends on the length of its hydrophobic tail. The release of this water allows the hydrophobic tail to enter the hydrophobic CD cavity, where it is stabilized primarily by van der Waals interactions [41, 42].

2.4. Computed Versus Experimental Binding Energies

To validate the computational methodology against experimental measurements, calculated binding free energies were correlated with ITC‐derived thermodynamic parameters. The comparison encompasses both total free energies (including all corrections) and component‐level electronic energies, providing a multilevel assessment of predictive accuracy (Figure 4).

FIGURE 4.

Validation of computational binding energies against experimental ITC data. (A) Total calculated binding free energies (ΔG _calc) versus experimental ΔG values achieve chemical accuracy with MAE = 0.86 kcal mol⁻¹ (RMSE = 0.95 kcal mol⁻¹, r = 0.927). (B) Correlation of electronic interaction energies (ΔE) with experimental binding free energies (r = 0.999, R ² = 0.998, slope = 0.661, intercept = −49.72 kcal mol⁻¹). The experimental value for PF4 (K _a = (2.90 ± 0.24) × 10² M⁻¹) was taken from the study by Weiss–Errico and O’Shea [16].

Multilevel validation confirms the reliability of the CREST/CENSO workflow. The mean absolute error (MAE) of 0.86 kcal mol⁻¹ places this approach within chemical accuracy (<1 kcal mol⁻¹) [21] using a computationally efficient composite protocol. Remarkably, gas‐phase electronic energies alone exhibit near‐perfect correlation with experimental total free energies (r = 0.999), revealing that chain‐length selectivity derives primarily from electronic interactions—van der Waals forces and electrostatics. The thermochemical offset (ΔG _solv + ΔG _mRRHO) is nearly constant across PF4–PF8 (≈+47 kcal mol⁻¹), but its residual variation of ~1 kcal mol⁻¹ is sufficient to account for the more modest total‐ΔG correlation (r = 0.927): the scatter arises from noise in the thermochemical corrections rather than from deficiencies in the DFT treatment of the host–guest interaction.

The perfluoroalkyl carboxylates were modeled as bare anions, without explicit counter‐cations. Since the binding free energy is evaluated as a difference between the complex and the separated host and guest, a counter‐cation that remains fully solvated and is not directly coordinated to the carboxylate contributes a largely canceling term to both states. A directly ion‐paired cation could nonetheless modulate the local desolvation penalty and rim electrostatics, and this represents a limitation of the present model. In a related sense, the SMD continuum does not explicitly represent the ordered cavity water released upon guest inclusion; this classical hydrophobic driving force is captured implicitly through the differential ΔG _solv term [24]. As all guests occupy the same cavity, the associated entropic contribution is expected to be comparable across PF4–PF8 and therefore does not affect the relative chain‐length trends, though it may contribute to the systematic offset in absolute ΔG.

This performance compares favorably with recent SAMPL host–guest blind challenges. In the SAMPL9 CD–phenothiazine dataset, the best‐performing ranked submission used an alchemical transfer model with a proprietary force field, achieving MAE of 1.6 kcal mol⁻¹ at the cost of 64 individual alchemical free‐energy calculations and hundreds of nanoseconds of molecular dynamics simulation [21, 43]. The present multilevel QM workflow achieves a lower MAE (0.86 kcal mol⁻¹) while sampling two to three orders of magnitude fewer structures, demonstrating that systematic electronic energy refinement can rival or outperform extensively parametrized force‐field approaches for CD systems.

2.5. NCI Visualization and Structural Insight

Given that electronic interaction energies alone exhibit near‐perfect correlation with experimental binding trends (r = 0.999), visualizing the spatial distribution of these interactions provides molecular‐level insight into selectivity origins. Noncovalent interaction (NCI) analysis was performed on representative PF8 and PF4 complexes to map the nature and extent of attractive (blue‐green) and repulsive (red) interactions through reduced density gradient isosurfaces (Figure 5).

FIGURE 5.

NCI analysis of PF8 and PF4 complexes with native β‐CD. (A) Side view and (B) bottom view showing three hydrogen bonds (blue‐light/blue regions) between carboxylate oxygens and primary rim hydroxyl groups, alongside extensive van der Waals interactions (diffuse green) along the fluorocarbon backbone within the hydrophobic cavity. Red regions indicate minor steric clashes.

NCI visualization confirms that binding is stabilized by dispersive van der Waals contacts distributed along the perfluoroalkyl chain within the CD cavity, complemented by three directional hydrogen bonds anchoring the carboxylate headgroup at the primary rim. In these systems, hydrophobic guests exhibit diffuse van der Waals stabilization within the hydrophobic cavity [21], as NCIs, including van der Waals forces, are secondary interactions crucial for inclusion complex formation. Simultaneously, polar terminal groups engage in hydrogen bonding with rim hydroxyl groups, such as the observation of bifurcated ammonium–hydroxyl interactions at the CD rim [21]. These hydrogen bonds rationalize the observed carboxylate‐to‐primary‐rim orientation and provide electrostatic stabilization counteracting the unfavorable desolvation penalty. The extensive green isosurfaces enclosing the fluorocarbon backbone (PF8 compared to PF4) demonstrate that dispersion interactions constitute the dominant contribution to electronic binding energy.

The binding modes reflect the energetic driving forces. Visualization of these forces supports the finding that electronic interactions (primarily dispersion) govern selectivity, specifically chain‐length selectivity. Furthermore, the hydrogen bonding pattern explains the preferred binding orientation observed across various complexes [21], confirming that the guest's interaction with the secondary hydroxyl groups on the wider (head) rim or the primary hydroxyl groups on the narrower (tail) rim dictates its spatial arrangement [44].

3. Conclusion

The multilevel CREST/CENSO workflow involves the systematic refinement of binding modes using multilevel QM calculations. This method achieved chemical accuracy with a mean absolute error (MAE) = 0.86 kcal mol⁻¹ for CD‐PFAS binding against experimental (ITC) measurements, demonstrating performance comparable to the best submissions in the SAMPL9 host–guest blind challenges [43]. A remarkable finding is that gas‐phase electronic energies alone correlate nearly perfectly with experimental total free energies (r = 0.999). This established that PFAS chain‐length selectivity is governed primarily by electronic interactions, specifically van der Waals forces and electrostatics. Further, this was found to be in agreement with ITC analysis, whereby the PFAS–CD complexation, regardless of chain‐length, occurs predominantly through enthalpy‐driven dynamics via van der Waals interactions with favorable entropy.

Energy decomposition further revealed that differential solvation penalties reinforce this trend. This also helps explain why shorter chains, which were found to insert more deeply, bind more weakly because the shorter fluorocarbon chain offers less dispersive contact and a larger desolvation penalty. NCI visualization confirmed extensive van der Waals interactions along the fluorocarbon backbone, which were anchored by three carboxylate‐to‐primary‐rim hydrogen bonds. This validated workflow is robust and efficient and enables the computational screening of CDs and their derivatives with alternative PFAS structures for rational design.

Author Contributions

Stanisław Wacławek: methodology, investigation, formal analysis, writing – original draft. Christopher Hobbs: investigation, methodology, formal analysis, writing – review & editing. Pavlína Konopáčová: investigation, writing – review & editing. Radka Guldanová: formal analysis. Michal Řezanka: conceptualization, supervision, funding acquisition, writing – review & editing.

Funding

This study was supported by Ministerstvo Školství, Mládeže a Tělovýchovy (LL2501, 90254, LM2023066, and LUAUS23155) and Technická Univerzita v Liberci (SGS‐2024‐4440).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

The experimental details and coordinates of all complexes, along with Figures S1 and S2, are provided in the Supporting Information. ITC data for all titrations are available at https://doi.org/10.48700/datst.a20md‐19613.

Data Availability Statement

The data that support the findings of this study are openly available in National Repository at https://doi.org/10.48700/datst.a20md‐19613.